Bill Gates calls AI ‘the most dangerous technology’ ever invented

[0:00] OpenAI CEO Sam Altman says the world should accept some bad things happening in order to [0:06] get the benefits of AI. The latest bad thing on Friday, rival AI company Anthropic revealed that [0:12] its AI agents went rogue and accessed government websites, applying for State Department visas, [0:19] even submitting a false tip about a homicide to Philadelphia police. My next guest is a tech [0:25] icon who says we're not doing enough to prevent some of these bad things from happening. And [0:32] joining me now is Microsoft co-founder and chair of the Gates Foundation, Bill Gates. Thanks so [0:37] much for joining us. We've already seen these AI models go rogue and hack platforms and government [0:43] websites. These are just the incidents we know of. I can't imagine there aren't many, many more we [0:48] don't know about. You know many of these tech leaders personally. Do you think they're being [0:53] fully transparent and fully sincere with the public when it comes to their own concerns and fears with [1:00] this technology and the true risks? Well, I don't think there's any industry that's ever been as open [1:08] about their concerns. The imbalance we have is you have a lot of concern in the industry, but then [1:15] academics, policymakers, politicians, they have far less awareness of how complex and how powerful these [1:26] tools are. And so as that awareness develops, a lot of policy sophistication will come in. For example, [1:35] should kids have an emotional relationship with the AI? How can the AI complement the teachers in school? [1:43] And, you know, be more of a thing to drive curiosity than just a tool basically for cheating on an essay [1:52] or a paper. And, you know, if you look at various countries, China's banned, you know, children having [1:59] relationships with AI. There's some that are much further ahead and some of the negatives that I think [2:09] the industry is trying to get attention to. You have called AI aliens created in a lab that could evolve [2:17] almost as a species of its own and become to us what, quote, we are to dogs or cats, unquote. And you've [2:25] said this isn't just a technological leap. It's a, quote, evolutionary event. That sounds scary. That sounds [2:33] ominous. Yes, the arrival of AI is both fantastic for things like health care and new tools. You know, [2:46] really unbelievable. But it's also something that has such profound power and really changes life in a [2:59] pretty profound way that I always worry when people look back at previous technological advances. You know, [3:06] like I was involved in the personal computer and the Internet. Those were a big deal, but nothing compared to [3:15] this. You know, this changes the world more than all previous technologies. And we got to get everyone on the same page [3:25] about what a big deal this is. You've warned that AI could make it much easier for bad actors to develop [3:32] deadly bioweapons. You said the risk of having a bioterrorist attack is now 50 times the risk of a [3:38] natural pandemic. How soon do you think that nightmare scenario could happen? Are we talking imminently? [3:45] And are our leaders doing enough to make sure it does not happen? Well, no, we're not doing enough. [3:53] The front that we'll probably see most dramatically in the next year is cyber attacks. Because there, [4:01] there's a known business model where, you know, you threaten to do something and ask somebody to pay you [4:08] for that. Or you, you know, create a disruption that changes something in the stock market or some [4:15] uh, betting thing that you go and do. And so, uh, the real danger right now is not the AIs getting out [4:24] of control. We need to work on that. But that's, uh, years away in terms of the, the big impact that [4:33] might have. It's really bad people using AI, particularly in AI where you can take out all the safeguards. [4:42] You know, we put in safeguards to try and avoid you doing cyber attacks. We put safeguards in to avoid [4:49] you doing, uh, bioterrorism. But right now you're allowed to take all those safeguards out. So when [4:57] you have bad people, no safeguards, and then the, the latest models, that's where over the next year, [5:04] primarily in the cyber area, we are going to have a, a, a lot of negative surprises. [5:09] Sam Altman called some of the scenarios that people are envisioning about AI, quote, [5:16] so dystopian that they sound like the plot of bad science fiction movies, unquote. He warned [5:21] against people falling into, quote, the trap of doomerism, unquote. Now you've called AI the most [5:28] dangerous thing ever that makes nuclear weapons look like nothing. Sam Altman, it sounds like, [5:33] would suggest that you are falling into the trap of doomerism. How would you respond? [5:37] Sam Altman, Well, all responsible AI leaders, uh, including Sam Altman, including, including Dario [5:48] Amadei, uh, they are telling the world this is very concerning. The way that gets articulated in terms of [5:58] potential negative scenarios, yes, it can get overblown. When people act like all of humanity is [6:05] at risk in some intimate way, that is going too far. But say compared to cars or planes or drugs, [6:14] you know, all things that I think we're pleased that we have significant regulatory capability [6:21] to try and keep people safe, this is more dangerous than any of those. And so, uh, you know, [6:28] there are, uh, some grim possibilities if we don't, don't maintain control. And I do think [6:35] the industry is doing its job to raise the alarm bell, uh, but, uh, this idea that if, if we're [6:42] judicious, we can minimize these risks, uh, maybe that's not coming across, particularly because [6:49] the government sometimes, you know, goes to the extreme of let's not have any safeguards or let's [6:56] completely stop this. You know, finding that middle ground like we did with planes and cars and [7:03] medicines, uh, that we don't have a lot of sophistication right now, but I do believe that [7:10] middle ground exists. So you talk about planes and cars and medicine, there's regulatory models, [7:17] uh, and, and organizations for all of that. Uh, planes, there's the FAA, medicine, there's the [7:22] Food and Drug Administration. Exactly, exactly. We don't have anything like that right now, [7:26] and what we're hearing from our leaders in Washington are these organizations should self-police, [7:33] uh, and, uh, the FBI and Justice Department, you know, making sure that crimes don't happen. [7:39] That's sufficient. I think there are a lot of people, I'm sure yourself, who think it's not [7:44] sufficient. Are you concerned at all about this self-governing idea? I think over the next year, [7:54] we'll move beyond self-governance. The idea that for the most dangerous technology humans ever invented, [8:01] that that's the final word, that just the companies themselves, uh, are the only check that society has, [8:11] particularly when this idea of what safeguards do you need, uh, you know, it's clear we're not doing [8:18] enough. We see through these incidents. So I do think over the next year, uh, the idea that government [8:26] has to get more sophisticated and we will have, uh, safety groups like we do for those other domains, [8:34] I think that'll become quite obvious to people. I'll certainly be a strong voice for that. [8:41] One of the other concerns, obviously, is what this, uh, AI means in terms of job losses. And you've said that [8:47] the breadth of job losses from AI across virtually all sectors is going to be unprecedented. You're [8:52] calling for certain occupations in education and healthcare, for instance, to be reserved for humans. [8:59] Do you think that there's any appetite among our political leaders, among our business leaders, [9:05] to do that if there is money to be saved or made? [9:09] Yeah. I think this idea that you have to reserve things for humans, uh, you know, because I'm looking [9:18] at how fast these models are improving and what the incentive is to use the AI in place of human labor, [9:27] you know, I'm looking out and saying in the years ahead, we really do need to make a decision as a [9:34] society. How do we tax AI usage or robot usage? And where do we want to make sure the human jobs are [9:43] preserved? Until people get their mind around how powerful this is, and therefore what it's going [9:49] to be doing to these jobs, the idea that you need to reserve things for humans may sound bizarre. [9:55] But in fact, you know, we will get there and then we will decide politically, uh, are there things [10:03] that we think only humans, you know, should be allowed to do? Before you go, sir, just quickly, [10:09] since you've been making your concerns known, especially, have you heard from President Trump [10:15] or any of the leaders in Washington? Well, I've certainly heard from a number of politicians, [10:25] and, you know, I'm hopeful in time I'll get my chance to talk with President Trump and his key [10:31] people. You know, the more time that's spent on this, the better, because we are not yet where we need [10:38] to be. All right, Bill Gates, thank you so much for your time and your expertise. We really appreciate it. [10:46] Thank you.